- Email: [email protected]
Do airlines always suffer from crashes?
Economics Letters 118 (2013) 113–117
Contents lists available at SciVerse ScienceDirect
Economics Letters journal homepage: www.elsevier.com/locate/ecolet
Do airlines always suffer from crashes? Jerry C. Ho ∗ , Mei Qiu, Xiaojun Tang School of Economics and Finance, Massey University, Auckland, New Zealand
article
info
Article history: Received 26 April 2012 Received in revised form 24 September 2012 Accepted 24 September 2012 Available online 5 October 2012
abstract We examine the impact of aviation disasters on the stock prices of the crash airlines and their rival airlines. Results show that the crash airlines experience deeper negative abnormal returns as the degree of fatality increases. The stock prices of the rival airlines also suffer in large-scale disasters but benefit from the disasters when the fatality is minor. © 2012 Elsevier B.V. All rights reserved.
JEL classification: G14 Keywords: Air crashes Stock market Contagion effect Switch effect Event study
1. Introduction This paper shows that the influence of an aviation disaster on the post-crash equity value of airlines, both crash and non-crash, is closely related to the level of fatality involved. Results show that air crash disasters with a more severe fatality – more people killed in an accident – have a larger negative impact on the stock prices of the crash airlines over the post-crash period. However, the stock prices of the non-crash airlines benefit from such disasters when the fatality is minor. The literature shows that air crashes reduce the market value of the equity of the crash airlines due to the loss of the aircraft and other costs incurred to handle the aftermath of crashes. Chance and Ferris (1987) show that the stock market responds instantly to air crash news. Borenstein and Zimmerman (1988) find that the shareholders of the crash airlines lose about one percent of their wealth. Air crashes adversely affect the stock prices of the crash airlines because they incur substantial financial losses during and after the disasters. Consequently, their stock prices plunge to reflect the decreases in the airlines’ expected future cash flows. In contrast, the direction of the impact of aviation disasters on the stock prices of the crash airline’s rivals depends on the
∗ Correspondence to: School of Economics and Finance, Massey University, Private Bag 102904, Auckland 0745, New Zealand. Tel.: +64 9 414 0800x9229; fax: +64 9 441 8177. E-mail address: [email protected] (J.C. Ho). 0165-1765/$ – see front matter © 2012 Elsevier B.V. All rights reserved. doi:10.1016/j.econlet.2012.09.031
interaction of the ‘‘contagion’’ effect and the ‘‘switch’’ effect. On one hand, the ‘‘contagion’’ effect arises when the tragic air crash news also influences the business of the non-crash airlines if it provokes the general public’s concern for air-travel safety, which results in a decline in the overall air travel demand (Gigerenzer, 2004; Silver et al., 2002; Sivak and Flannagan, 2004). Furthermore, after air crashes investors may anticipate upward industry-wide insurance rate adjustments (Mitchell and Maloney, 1989) and tightened industrial regulations (Chance and Ferris, 1987). Such increased operating costs reduce expected cash flows. On the other hand, the ‘‘switch’’ effect occurs when air travelers fly with the competitors of the crash airlines. The misfortune of the crash airlines benefits the non-crash airlines. Bosch et al. (1998) show that the ‘‘switch’’ effect (the ‘‘contagion’’ effect) is present for the non-crash airlines which have high (low) market overlaps with the crash airlines. Our study relates the degree of severity of aviation disasters to the post-event stock prices of the airlines. For single-digit fatality disasters, the shareholders of the crash airlines suffer immediate wealth losses in the first post-crash week while the shareholders of the non-crash airlines consistently enjoy wealth gains over the entire post-event period, implying the dominance of the ‘‘switch’’ effect over the ‘‘contagion’’ effect. When the death toll climbs to a double-digit or a triple-digit number, all airlines experience significant reductions in equity value, indicating the ‘‘contagion’’ effect of the large-scale disasters within the entire airline industry. Higher degrees of fatality lead to deeper reductions in the market value of their equity.
114
J.C. Ho et al. / Economics Letters 118 (2013) 113–117
Table 1 Post-crash abnormal returns. Date
AR
Panel A: crash airlines (n = 113) 0 −1.58 +1 −0.32 +2 −0.01 +3 −0.04 +4 0.01 +5 −0.46 +10 −0.02 +15 0.17 +20 −0.39 +25 −0.01
Patell test
Sign test
Window
CAR
Patell test
Sign test
−6.658**** −1.468
−4.151**** −0.947
0.157 0.114 −0.621 −1.959* −0.038 0.423 −1.886* 0.601
−0.193 −2.078** −0.570 −0.184 −0.758 −1.324
(0, 0) (0, 1) (0, 2) (0, 3) (0, 4) (0, 5) (0, 10) (0, 15) (0, 20) (0, 25)
−1.58 −1.90 −1.91 −1.95 −1.94 −2.41 −1.85 −2.13 −2.74 −2.94
−6.658**** −5.752**** −4.600**** −3.923**** −3.753**** −4.191**** −2.955*** −2.788*** −3.121*** −3.074***
−4.151**** −2.832*** −3.774**** −3.020*** −2.643*** −3.020*** −1.323 −2.265** −2.265** −0.946
(0, 0) (0, 1) (0, 2) (0, 3) (0, 4) (0, 5) (0, 10) (0, 15) (0, 20) (0, 25)
−0.13 −0.24 −0.26 −0.16 −0.26 −0.11 −0.03 −0.31 −0.36
−0.472 −1.515 −2.254** −2.061** −1.413 −2.058** −1.295 −1.197 −1.766* −1.511
−0.027 −0.780 −1.590 −1.764* −1.706* −1.822* −0.259 −1.590 −0.896 −0.086
Panel B: non-crash airlines (n = 1, 199) 0 0.01 −0.472 +1 −0.14 −1.668* +2 −0.11 −1.760* +3 −0.02 −0.262 +4 0.10 0.858 +5 −0.10 −1.887* +10 0.02 −0.078 +15 0.20 2.239** +20 −0.04 −0.961 +25 −0.06 −0.288
0.184
1.315
−0.027 −0.780 −2.343** −0.664 −0.606 −0.838 −1.100 2.866***
−0.548 −1.937*
0.01
AR is average abnormal return and CAR stands for cumulative average abnormal return. Figures are reported in percentage rate terms. * Statistical significances at the 10% level. ** Statistical significances at the 5% level. *** Statistical significances at the 1% level. **** Statistical significances at the 0.1% level.
2. Data and methodology We obtained the information on the U.S. air crashes that happened over the 1950–2009 period from the ASN Aviation Safety Database.1 The airlines must meet the following criteria to be included in our sample. First, they must be U.S.-based companies listed on the NYSE, AMEX, or Nasdaq at the time of crash. Second, the company must have at least 140 pre-event daily prices available in the CRSP (The Center for Research in Security Prices) database. Third, only the crashes that have caused the death of either the aircraft occupants or people on the ground are investigated. To avoid confounding effects, our analysis eliminates the crash airlines with other significant corporate events within five days after a crash.2 Our final sample includes 113 air crash disasters and 1199 non-crash airlines. We classify all events into three categories based on the degree of severity of a disaster measured by the number of digits in the fatality figures. Inspired by the psychology literature, we use 10 and 100 as the cutoff fatality rates. Mitchell (2001) claims that the use of the decimal system suggests a natural tendency for people to think in terms of 10s and powers of 10. Grouping by the powers of 10 facilitates the mental order and provides computationally convenient breaks between groups. Increased psychological significance occurs at 10s and increasing powers of 10. Apart from reporting the actual numbers of deaths involved in accidents or disasters, the news media often describes the seriousness of events in terms of the number of digits of the death toll. Therefore, we use 10 and 100 as the cutoff fatality rates which may reflect the psychological levels of the public’s perceived seriousness of fatality for the events. Our low-fatality group consists of disasters with single-digit numbers of fatality and
1 Available at http://aviation-safety.net 2 We search in Factiva announcements of dividend payment, revenue or earning surprise, merger acquisition, new services, winning of major government contracts, change of key management personnel and filing of law suit for large damages within five days after a crash.
involves 44 crash airlines and 465 non-crash airlines. The mediumfatality group includes the disasters that kill 10–99 people and affects 56 crash airlines and 600 non-crash airlines. The highfatality group contains disasters with three-digit numbers of fatality and involves 13 crash airlines and 134 non-crash airlines. The date that an air crash occurred is defined as the event day (day zero). If a crash happened on a non-trading day then the following trading day is used as the event day. The market-model is estimated over the time interval from 255 through 46 trading days prior to the event day. The abnormal return (ARit ) for airline i as of event date t is defined as: ARit = Rit − ( αi + βi Rmt ),
(1)
where Rit is the actual stock return of the ith airline on day t, and Rmt is the return on the S&P 500 stock market index on day t. The average abnormal return (AARt ) of a group of airline companies is calculated as: N
AARt =
ARit
i=1
N
.
(2)
The cumulative average abnormal return (CART1 T2 ) over an event window between days T1 and T2 is: T2 N
CART1 T2 =
ARi
i =1 t =T 1
N
.
(3)
3. Empirical results Table 1 shows that both the crash and the non-crash airlines experience negative post-crash abnormal stock returns. A noticeable difference is that the stock prices of the crash airlines immediately react adversely to the disasters from the event date while the non-crash airlines begin to respond one day later. Panel A shows that the crash airlines lose 1.58% of their market value on the
J.C. Ho et al. / Economics Letters 118 (2013) 113–117
115
Table 2 Abnormal returns of crash airlines by fatality. Date
AR
Panel A: low fatalities (n = 44) 0 −0.85 +1 0.28 +2 −0.18 +3 0.00 +4 0.58 +5 −0.36 +10 −0.25 +15 −0.37 +20 −0.40 +25 0.97 Panel B: medium fatalities (n = 56) 0 −1.40 +1 −0.35 +2 0.01 +3 0.11 +4 −0.28 +5 −0.02 +10 0.19 +15 0.57 +20 −0.41 +25 −0.50 Panel C: high fatalities (n = 13) 0 −4.89 +1 −2.19 +2 0.47 +3 −0.84 +4 −0.69 +5 −2.69 +10 −0.11 +15 0.26 +20 −0.31 +25 −1.19
Patell test
Sign test
Window
CAR
Patell test
Sign test
−2.394**
−1.988**
0.488 0.201 −0.087 0.832 −1.071 −0.075 −0.771 −1.276 2.957***
0.732 0.732 0.127 −0.175 0.127 1.034 −1.081 −0.779 3.451****
(0, 0) (0, 1) (0, 2) (0, 3) (0, 4) (0, 5) (0, 10) (0, 15) (0, 20) (0, 25)
−0.85 −0.57 −0.75 −0.75 −0.17 −0.53
−2.394** −1.360 −0.975 −0.867 −0.405 −0.811
−1.988** −0.477 −1.988** −1.081 −0.477 −0.175
0.431 0.131 0.155 0.022
0.429 0.429 1.034 0.429
−4.829**** −1.097 −0.429
−3.322**** −0.913 −0.378 −0.378 −1.984**
(0 , 0 ) (0 , 1 ) (0 , 2 ) (0 , 3 ) (0 , 4 ) (0 , 5 ) (0, 10) (0, 15) (0, 20) (0, 25)
−1.40 −1.75 −1.74 −1.62 −1.90 −1.92 −2.39 −2.27 −3.40 −2.97
−4.829**** −4.184**** −3.674**** −2.962*** −3.146*** −3.041*** −3.063*** −2.434** −2.958*** −2.700***
−3.322**** −2.787*** −2.519** −2.519** −2.251** −3.054*** −1.447 −2.786*** −3.589**** −1.447
(0 , 0 ) (0 , 1 ) (0 , 2 ) (0 , 3 ) (0 , 4 ) (0 , 5 ) (0, 10) (0, 15) (0, 20) (0, 25)
−4.89 −7.08 −6.61 −7.45 −8.14 −10.83 −8.45 −10.00 −11.45 −14.37
−5.202**** −5.773**** −4.144**** −3.823**** −3.792**** −4.553**** −3.148*** −3.408**** −3.349**** −3.500****
−1.687* −1.687* −2.243** −1.687* −2.243** −2.243** −1.687* −1.687* −1.131 −0.574
0.473
−1.191 −0.526
0.158
0.075 1.314 −1.335 −1.558
−0.644 −0.109 −1.180 −1.715*
−5.202**** −2.948***
−1.687* −2.243** −0.018 −0.018 −1.687* −2.243** −0.018 −0.018 −0.018
0.982
−0.485 −0.890 −2.715*** −0.131 −0.060 −0.443 −0.437
1.094
0.79 0.38 0.69 0.49
Crashes leading to less than 10 fatalities are included in Low Fatalities. Crashes that have caused between 10 and 99 deaths are classified as Medium Fatalities. The high Fatalities group contains crashes that have killed more than 100 people. AR is the average daily abnormal return of all airline companies in the same group. CAR is the cumulative average abnormal return of the event window of all airlines in the same sample group. Figures are reported in percentage rate terms. * Statistical significances at the 10% level. ** Statistical significances at the 5% level. *** Statistical significances at the 1% level. **** Statistical significances at the 0.1% level.
Fig. 1. Post-crash CARs crash (solid line) vs. non-crash airlines (dotted line).
crash date and this loss is statistically significant. In contrast, Panel B shows that the non-crash airlines do not experience any value loss in their shares on the crash date. Fig. 1 demonstrates their CARs over the 25-day post-event window. Apparently, despite the wealth of the shareholders of both groups suffer following the aviation disasters, the shareholders of the crash airlines experience a larger short-term wealth loss than those of the non-crash airlines. Kaplanski and Levy (2010) claim that aviation disasters could cause negative sentiment due to bad moods and anxieties and may in turn affect the stock prices of airlines. We argue that the fatality reinforces such a negative atmosphere and hence the scale of fatality plays a critical role in determining the post-crash stock price movement of the airlines. Table 2 shows that the investors of the crash airlines suffer larger losses on the crash day as the number of people killed in the accidents increases. Specifically, the average abnormal return of the crash airlines on the event date for the single-digit disasters (Panel A) is as low as −0.85%
Fig. 2. Degree of fatality and cumulative abnormal returns to crash airlines.
compared to −1.40% for the double-digit disasters (Panel B) and −4.89% for the triple-digit disasters (Panel C). Fig. 2 reveals that the crashes with larger fatality cause greater losses of the equity value of the crash airlines. The negative post-crash cumulative return drops dramatically as time elapses for the disasters with tripledigit fatality but only mildly for the double-digit fatality events. Interestingly, the (cumulative) negative reactions to single-digit fatality crashes disappear one week after the events,3 suggesting that the market soon recovers its confidence in the crash airlines after a week if the fatality is minor.
3 The authors thank for the anonymous referee’s suggestion to explore the patterns in CARs in a longer post-event window.
116
J.C. Ho et al. / Economics Letters 118 (2013) 113–117
Table 3 Abnormal returns of non-crash airlines by fatality. Date
AR
Panel A: low fatalities (n = 465) 0 0.13 +1 0.11 +2 0.04 +3 0.05 +4 0.17 +5 0.00 +10 0.40 +15 0.05 +20 0.02 +25 0.09 Panel B: medium fatalities (n = 600) 0 −0.01 +1 −0.21 +2 −0.30 +3 −0.06 +4 0.05 +5 −0.20 +10 −0.22 +15 0.30 +20 −0.06 +25 −0.28 Panel C: high fatalities (n = 134) 0 −0.34 +1 −0.68 +2 0.28 +3 −0.11 +4 0.06 +5 0.00 +10 −0.28 +15 0.31 +20 −0.12 +25 0.37
Patell test
Sign test
Window
0.649 0.894 0.514 0.032 1.231 −0.169 2.466** −0.677 −0.190 1.179
1.232 0.210 −0.441 −0.441 −0.906 0.767 1.650* 1.512 0.675 0.303
(0, 0) (0, 1) (0, 2) (0, 3) (0, 4) (0, 5) (0, 10) (0, 15) (0, 20) (0, 25)
0.13 0.24 0.27 0.33 0.50 0.50 1.27 1.56 1.53 1.25
−0.515 −2.075** −3.223*** −0.205
−0.337
(0, 0) (0, 1) (0, 2) (0, 3) (0, 4) (0, 5) (0, 10) (0, 15) (0, 20) (0, 25)
−0.01 −0.22 −0.52 −0.58 −0.54 −0.74 −1.08 −0.98 −1.45 −1.38
−0.515 −1.824* −3.360**** −3.028*** −2.585*** −3.388**** −3.531**** −2.725**** −3.377**** −2.796***
−0.337
(0, 0) (0, 1) (0, 2) (0, 3) (0, 4) (0, 5) (0, 10) (0, 15) (0, 20) (0, 25)
−0.34 −1.01 −0.73 −0.84 −0.77 −0.78 −0.57 −1.36 −1.62 −1.35
−1.536 −2.701*** −1.860* −1.795* −1.640 −1.448 −0.909 −1.616 −1.614 −0.930
−1.664* −2.876*** −1.491 −1.838* −2.011** −0.799 −0.106 −0.626 −1.318 −1.491
−0.741 −2.376**
0.154 −3.119*** 0.236 0.563 −2.055** −1.974** 1.626 −1.238 −3.447****
−1.536 −2.271**
−1.664* −3.050***
0.160
−2.568** −1.859* 3.220***
0.593
0.413
−0.412 −0.066
−1.664* −1.318
0.102
−0.888 1.149
0.413
−2.184** 2.318**
−0.952
−0.279
1.945*
0.933
CAR
Patell test
Sign test
0.649 1.089 1.196 1.094 1.548 1.322 2.416** 2.037** 1.864* 1.246
1.232 0.024 0.582 0.210 −0.069 1.418 2.906*** 1.605 2.348** 1.976**
0.236
−2.055** −1.810* −1.401 −3.446**** −2.874*** −3.365**** −2.710*** −1.156
Crashes leading to fewer than 10 fatalities are included in Low Fatalities. Crashes that have caused between 10 and 99 deaths are classified as Medium Fatalities. The High Fatalities group contains crashes that have killed more than 100 people. AR is the average daily abnormal return of all airline companies in the same group. CAR is the cumulative average abnormal return of the event window of all airlines in the same sample group. Figures are reported in percentage rate terms. * Statistical significances at the 10% level. ** Statistical significances at the 5% level. *** Statistical significances at the 1% level. **** Statistical significances at the 0.1% level.
Fig. 3. Degree of fatality and cumulative abnormal returns to non-crash airlines.
Fig. 3 shows that the direction of the impact of the aviation disasters on the stock prices of the non-crash airlines is determined by the relative importance of the ‘‘switch’’ effect and the ‘‘contagion’’ effect. We expect that the ‘‘switch’’ effect dominates the ‘‘contagion’’ effect when the fatality is minor, while the ‘‘contagion’’ effect begins to dominate the ‘‘switch’’ effect as the fatality increases. In the former scenario, the non-crash airlines may benefit from the misfortune of the crash airlines because passengers may switch to those non-crash airlines following aviation disasters. In the later scenario where large-scale fatality occurs, people may just avoid air travel after the disasters. As a result, the tragic news creates a contagion of fear that spreads through the entire airline industry.
Table 3 presents our test of this hypothesis. Panel A shows that both the ARs and CARs for the low-fatality air crashes during the post-crash period are consistently positive, suggesting that investors perceive that a small-fatality accident would benefit the non-crash airlines because passengers of the crash airlines may switch to the non-crash airlines and hence increase the cash flows of the rivals of the crash airlines. On the contrary, as the number of people killed exceeds ten, investors predict that such relatively large tragedies would cause a comprehensive bad sentiment concerning air travel safety among the public and potential increased operating costs for the entire airline industry. Consequently, even the non-crash airlines are adversely affected by the misfortune of the crash airlines. The consistently negative CARs of the non-crash airline reported in Panels B and C clearly provide supportive evidence for our argument that the ‘‘contagion’’ effect prevails in large-scale disasters. Overall, consistent with the findings of Kaplanski and Levy (2010) that large-scale aviation disasters affect people’s moods and increase their anxieties and consequently reduce their investment in risky assets, our results indicate that the more disastrous the aviation accidents, the stronger the fears and anxieties about air travel that impact the stock market. Finally, we examine whether there exists a difference between the means of the post-crash abnormal cumulative returns between each pair of fatality groups of stocks. Table 4 shows that the postcrash stock performance of both the crash and non-crash airlines is more significantly negatively affected by the high-fatality disasters
J.C. Ho et al. / Economics Letters 118 (2013) 113–117 Table 4 Comparisons of post-crash abnormal returns between different fatality groups. Event window
Low–medium
Medium–high
Panel A: comparisons between crash airline groups (0 , 0 ) 0.85 1.32 (0 , 1 ) 2.13** 2.44** (0 , 2 ) 3.07*** 0.49 (0, 3) 3.09*** 0.55 (0, 4) 3.05*** 0.46 (0, 5) 3.39*** 0.07 (0, 10) 3.17* 6.06* (0, 15) 2.90 8.13* (0, 20) 4.09 8.05 (0, 25) 2.79 11.34 Panel B: comparisons between non-crash airline groups (0, 0) 1.06 1.62 (0, 1) 1.68* 2.32** (0, 2) 1.17 2.06** (0, 3) 0.83 1.91* (0, 4) 1.49 2.09** (0, 5) 1.19 2.14** (0, 10) 2.35*** −0.52 (0, 15) 2.45*** 0.32 (0, 20) 2.98*** 0.18 *** (0, 25) 2.47 −0.46
Low–high 1.81* 3.70*** 2.23** 2.43** 2.34** 2.22** 9.23** 11.03** 12.13** 14.13* 1.85* 2.83*** 2.44** 2.20** 2.67*** 2.49** 1.84** 2.77*** 3.16*** 2.01
This table reports whether two fatality groups have an equal mean of cumulative abnormal return in an event-window. For both the crash airline and the non-crash airline sample, Low, Medium, and High contain all observations relating to air crashes that have killed fewer than 10 people, between 10 and 99 people and over 100 people, respectively. The t-statistics of the tests are reported followed by the indications of statistical significance levels. * Statistical significances at the 10% level. ** Statistical significances at the 5% level. *** Statistical significances at the 1% level.
than by the low-fatality accidents. Similar findings, though less evident, also exist between the low- and medium- fatality groups as well as the medium- and high-fatality groups. These results indicate that the market responds asymmetrically to the aviation disasters with various degrees of fatality. 4. Concluding remarks We examine the stock price performance of the airline industry following aviation disasters that have caused different levels of fatality. The results show that stock investors perceive differently the degree of the severity that aviation disasters may influence the stock performance of the crash and non-crash airlines. The crash airlines suffer larger losses of the equity value than
117
their competitors when aviation disasters occur. When aviation disasters result in single-digit fatality, the stock prices of the crash airlines initially respond negatively to the tragedy but recover quickly a week later. A higher degree of fatality has a larger and more persistent negative impact on the stock prices of the crash airlines following air crashes. Our explanation is that aviation disasters cause considerable financial damage to the crash airlines due to the reduction in air travel demand, possible future higher insurance costs, reputation and brand name damage, and customer goodwill loss. For the non-crash airlines, the ‘‘switch’’ effect dominates if the number of deaths in an accident is lower than 10 as these airlines benefit from the misfortune of the crash airlines. At higher fatality levels, however, the ‘‘contagion’’ effect dominates, as suggested by the declines of the stock prices following more serious crashes. Our explanation for the dominance of the ‘‘switch’’ effect for smallscale aviation disasters is that low fatality crashes receive less attention from the public and concerns about air travel safety. Air travelers may simply switch to the non-crash airlines to meet their travel needs after the occurrence of the disasters. On the contrary, large-scale air crashes tend to attract more attention from the media. Intensive and constant follow-up reports in the media for such devastating disasters reinforce greater fears and anxieties from the public than less serious ones. As a result, the non-crash airlines also suffer as the bad news creates a contagion of fear of air travelling that spreads through the public. References Borenstein, S., Zimmerman, M.B., 1988. Market incentives for safe commercial airline operation. The American Economic Review 78 (5), 913–935. Bosch, J.C., Eckard, E.W., Singal, V., 1998. The competitive impact of air crashes: stock market evidence. Journal of Law and Economics 41 (2), 503–519. Chance, D.M., Ferris, S.P., 1987. The effect of aviation disasters on the air transport industry. Journal of Transport Economics and Policy 31 (2), 151–165. Gigerenzer, G., 2004. Dread risk, September 11, and fatal traffic accidents. Psychological Science 15 (4), 286–287. Kaplanski, G., Levy, H., 2010. Sentiment and stock prices: the case of aviation disasters. Journal of Financial Economics 95 (2), 174–201. Mitchell, J., 2001. Clustering and psychological barriers: the importance of numbers. The Journal of Futures Markets 21 (5), 395–428. Mitchell, M.L., Maloney, M.T., 1989. Crisis in the cockpit? The role of market forces in promoting air safety. Journal of Law and Economics 32 (2), 329–355. Silver, R.C., Holman, E.A., McIntosh, D.N., Poulin, M., Gil-Rivas, V., 2002. Nationwide longitudinal study of psychological responses to September 11. The Journal of the American Medical Association 288 (10), 1235–1244. Sivak, M., Flannagan, M.J., 2004. Consequences for road traffic fatalities of the reduction in flying following September 11, 2001. Transportation Research 7 (4–5), 301–305.
Contents lists available at SciVerse ScienceDirect
Economics Letters journal homepage: www.elsevier.com/locate/ecolet
Do airlines always suffer from crashes? Jerry C. Ho ∗ , Mei Qiu, Xiaojun Tang School of Economics and Finance, Massey University, Auckland, New Zealand
article
info
Article history: Received 26 April 2012 Received in revised form 24 September 2012 Accepted 24 September 2012 Available online 5 October 2012
abstract We examine the impact of aviation disasters on the stock prices of the crash airlines and their rival airlines. Results show that the crash airlines experience deeper negative abnormal returns as the degree of fatality increases. The stock prices of the rival airlines also suffer in large-scale disasters but benefit from the disasters when the fatality is minor. © 2012 Elsevier B.V. All rights reserved.
JEL classification: G14 Keywords: Air crashes Stock market Contagion effect Switch effect Event study
1. Introduction This paper shows that the influence of an aviation disaster on the post-crash equity value of airlines, both crash and non-crash, is closely related to the level of fatality involved. Results show that air crash disasters with a more severe fatality – more people killed in an accident – have a larger negative impact on the stock prices of the crash airlines over the post-crash period. However, the stock prices of the non-crash airlines benefit from such disasters when the fatality is minor. The literature shows that air crashes reduce the market value of the equity of the crash airlines due to the loss of the aircraft and other costs incurred to handle the aftermath of crashes. Chance and Ferris (1987) show that the stock market responds instantly to air crash news. Borenstein and Zimmerman (1988) find that the shareholders of the crash airlines lose about one percent of their wealth. Air crashes adversely affect the stock prices of the crash airlines because they incur substantial financial losses during and after the disasters. Consequently, their stock prices plunge to reflect the decreases in the airlines’ expected future cash flows. In contrast, the direction of the impact of aviation disasters on the stock prices of the crash airline’s rivals depends on the
∗ Correspondence to: School of Economics and Finance, Massey University, Private Bag 102904, Auckland 0745, New Zealand. Tel.: +64 9 414 0800x9229; fax: +64 9 441 8177. E-mail address: [email protected] (J.C. Ho). 0165-1765/$ – see front matter © 2012 Elsevier B.V. All rights reserved. doi:10.1016/j.econlet.2012.09.031
interaction of the ‘‘contagion’’ effect and the ‘‘switch’’ effect. On one hand, the ‘‘contagion’’ effect arises when the tragic air crash news also influences the business of the non-crash airlines if it provokes the general public’s concern for air-travel safety, which results in a decline in the overall air travel demand (Gigerenzer, 2004; Silver et al., 2002; Sivak and Flannagan, 2004). Furthermore, after air crashes investors may anticipate upward industry-wide insurance rate adjustments (Mitchell and Maloney, 1989) and tightened industrial regulations (Chance and Ferris, 1987). Such increased operating costs reduce expected cash flows. On the other hand, the ‘‘switch’’ effect occurs when air travelers fly with the competitors of the crash airlines. The misfortune of the crash airlines benefits the non-crash airlines. Bosch et al. (1998) show that the ‘‘switch’’ effect (the ‘‘contagion’’ effect) is present for the non-crash airlines which have high (low) market overlaps with the crash airlines. Our study relates the degree of severity of aviation disasters to the post-event stock prices of the airlines. For single-digit fatality disasters, the shareholders of the crash airlines suffer immediate wealth losses in the first post-crash week while the shareholders of the non-crash airlines consistently enjoy wealth gains over the entire post-event period, implying the dominance of the ‘‘switch’’ effect over the ‘‘contagion’’ effect. When the death toll climbs to a double-digit or a triple-digit number, all airlines experience significant reductions in equity value, indicating the ‘‘contagion’’ effect of the large-scale disasters within the entire airline industry. Higher degrees of fatality lead to deeper reductions in the market value of their equity.
114
J.C. Ho et al. / Economics Letters 118 (2013) 113–117
Table 1 Post-crash abnormal returns. Date
AR
Panel A: crash airlines (n = 113) 0 −1.58 +1 −0.32 +2 −0.01 +3 −0.04 +4 0.01 +5 −0.46 +10 −0.02 +15 0.17 +20 −0.39 +25 −0.01
Patell test
Sign test
Window
CAR
Patell test
Sign test
−6.658**** −1.468
−4.151**** −0.947
0.157 0.114 −0.621 −1.959* −0.038 0.423 −1.886* 0.601
−0.193 −2.078** −0.570 −0.184 −0.758 −1.324
(0, 0) (0, 1) (0, 2) (0, 3) (0, 4) (0, 5) (0, 10) (0, 15) (0, 20) (0, 25)
−1.58 −1.90 −1.91 −1.95 −1.94 −2.41 −1.85 −2.13 −2.74 −2.94
−6.658**** −5.752**** −4.600**** −3.923**** −3.753**** −4.191**** −2.955*** −2.788*** −3.121*** −3.074***
−4.151**** −2.832*** −3.774**** −3.020*** −2.643*** −3.020*** −1.323 −2.265** −2.265** −0.946
(0, 0) (0, 1) (0, 2) (0, 3) (0, 4) (0, 5) (0, 10) (0, 15) (0, 20) (0, 25)
−0.13 −0.24 −0.26 −0.16 −0.26 −0.11 −0.03 −0.31 −0.36
−0.472 −1.515 −2.254** −2.061** −1.413 −2.058** −1.295 −1.197 −1.766* −1.511
−0.027 −0.780 −1.590 −1.764* −1.706* −1.822* −0.259 −1.590 −0.896 −0.086
Panel B: non-crash airlines (n = 1, 199) 0 0.01 −0.472 +1 −0.14 −1.668* +2 −0.11 −1.760* +3 −0.02 −0.262 +4 0.10 0.858 +5 −0.10 −1.887* +10 0.02 −0.078 +15 0.20 2.239** +20 −0.04 −0.961 +25 −0.06 −0.288
0.184
1.315
−0.027 −0.780 −2.343** −0.664 −0.606 −0.838 −1.100 2.866***
−0.548 −1.937*
0.01
AR is average abnormal return and CAR stands for cumulative average abnormal return. Figures are reported in percentage rate terms. * Statistical significances at the 10% level. ** Statistical significances at the 5% level. *** Statistical significances at the 1% level. **** Statistical significances at the 0.1% level.
2. Data and methodology We obtained the information on the U.S. air crashes that happened over the 1950–2009 period from the ASN Aviation Safety Database.1 The airlines must meet the following criteria to be included in our sample. First, they must be U.S.-based companies listed on the NYSE, AMEX, or Nasdaq at the time of crash. Second, the company must have at least 140 pre-event daily prices available in the CRSP (The Center for Research in Security Prices) database. Third, only the crashes that have caused the death of either the aircraft occupants or people on the ground are investigated. To avoid confounding effects, our analysis eliminates the crash airlines with other significant corporate events within five days after a crash.2 Our final sample includes 113 air crash disasters and 1199 non-crash airlines. We classify all events into three categories based on the degree of severity of a disaster measured by the number of digits in the fatality figures. Inspired by the psychology literature, we use 10 and 100 as the cutoff fatality rates. Mitchell (2001) claims that the use of the decimal system suggests a natural tendency for people to think in terms of 10s and powers of 10. Grouping by the powers of 10 facilitates the mental order and provides computationally convenient breaks between groups. Increased psychological significance occurs at 10s and increasing powers of 10. Apart from reporting the actual numbers of deaths involved in accidents or disasters, the news media often describes the seriousness of events in terms of the number of digits of the death toll. Therefore, we use 10 and 100 as the cutoff fatality rates which may reflect the psychological levels of the public’s perceived seriousness of fatality for the events. Our low-fatality group consists of disasters with single-digit numbers of fatality and
1 Available at http://aviation-safety.net 2 We search in Factiva announcements of dividend payment, revenue or earning surprise, merger acquisition, new services, winning of major government contracts, change of key management personnel and filing of law suit for large damages within five days after a crash.
involves 44 crash airlines and 465 non-crash airlines. The mediumfatality group includes the disasters that kill 10–99 people and affects 56 crash airlines and 600 non-crash airlines. The highfatality group contains disasters with three-digit numbers of fatality and involves 13 crash airlines and 134 non-crash airlines. The date that an air crash occurred is defined as the event day (day zero). If a crash happened on a non-trading day then the following trading day is used as the event day. The market-model is estimated over the time interval from 255 through 46 trading days prior to the event day. The abnormal return (ARit ) for airline i as of event date t is defined as: ARit = Rit − ( αi + βi Rmt ),
(1)
where Rit is the actual stock return of the ith airline on day t, and Rmt is the return on the S&P 500 stock market index on day t. The average abnormal return (AARt ) of a group of airline companies is calculated as: N
AARt =
ARit
i=1
N
.
(2)
The cumulative average abnormal return (CART1 T2 ) over an event window between days T1 and T2 is: T2 N
CART1 T2 =
ARi
i =1 t =T 1
N
.
(3)
3. Empirical results Table 1 shows that both the crash and the non-crash airlines experience negative post-crash abnormal stock returns. A noticeable difference is that the stock prices of the crash airlines immediately react adversely to the disasters from the event date while the non-crash airlines begin to respond one day later. Panel A shows that the crash airlines lose 1.58% of their market value on the
J.C. Ho et al. / Economics Letters 118 (2013) 113–117
115
Table 2 Abnormal returns of crash airlines by fatality. Date
AR
Panel A: low fatalities (n = 44) 0 −0.85 +1 0.28 +2 −0.18 +3 0.00 +4 0.58 +5 −0.36 +10 −0.25 +15 −0.37 +20 −0.40 +25 0.97 Panel B: medium fatalities (n = 56) 0 −1.40 +1 −0.35 +2 0.01 +3 0.11 +4 −0.28 +5 −0.02 +10 0.19 +15 0.57 +20 −0.41 +25 −0.50 Panel C: high fatalities (n = 13) 0 −4.89 +1 −2.19 +2 0.47 +3 −0.84 +4 −0.69 +5 −2.69 +10 −0.11 +15 0.26 +20 −0.31 +25 −1.19
Patell test
Sign test
Window
CAR
Patell test
Sign test
−2.394**
−1.988**
0.488 0.201 −0.087 0.832 −1.071 −0.075 −0.771 −1.276 2.957***
0.732 0.732 0.127 −0.175 0.127 1.034 −1.081 −0.779 3.451****
(0, 0) (0, 1) (0, 2) (0, 3) (0, 4) (0, 5) (0, 10) (0, 15) (0, 20) (0, 25)
−0.85 −0.57 −0.75 −0.75 −0.17 −0.53
−2.394** −1.360 −0.975 −0.867 −0.405 −0.811
−1.988** −0.477 −1.988** −1.081 −0.477 −0.175
0.431 0.131 0.155 0.022
0.429 0.429 1.034 0.429
−4.829**** −1.097 −0.429
−3.322**** −0.913 −0.378 −0.378 −1.984**
(0 , 0 ) (0 , 1 ) (0 , 2 ) (0 , 3 ) (0 , 4 ) (0 , 5 ) (0, 10) (0, 15) (0, 20) (0, 25)
−1.40 −1.75 −1.74 −1.62 −1.90 −1.92 −2.39 −2.27 −3.40 −2.97
−4.829**** −4.184**** −3.674**** −2.962*** −3.146*** −3.041*** −3.063*** −2.434** −2.958*** −2.700***
−3.322**** −2.787*** −2.519** −2.519** −2.251** −3.054*** −1.447 −2.786*** −3.589**** −1.447
(0 , 0 ) (0 , 1 ) (0 , 2 ) (0 , 3 ) (0 , 4 ) (0 , 5 ) (0, 10) (0, 15) (0, 20) (0, 25)
−4.89 −7.08 −6.61 −7.45 −8.14 −10.83 −8.45 −10.00 −11.45 −14.37
−5.202**** −5.773**** −4.144**** −3.823**** −3.792**** −4.553**** −3.148*** −3.408**** −3.349**** −3.500****
−1.687* −1.687* −2.243** −1.687* −2.243** −2.243** −1.687* −1.687* −1.131 −0.574
0.473
−1.191 −0.526
0.158
0.075 1.314 −1.335 −1.558
−0.644 −0.109 −1.180 −1.715*
−5.202**** −2.948***
−1.687* −2.243** −0.018 −0.018 −1.687* −2.243** −0.018 −0.018 −0.018
0.982
−0.485 −0.890 −2.715*** −0.131 −0.060 −0.443 −0.437
1.094
0.79 0.38 0.69 0.49
Crashes leading to less than 10 fatalities are included in Low Fatalities. Crashes that have caused between 10 and 99 deaths are classified as Medium Fatalities. The high Fatalities group contains crashes that have killed more than 100 people. AR is the average daily abnormal return of all airline companies in the same group. CAR is the cumulative average abnormal return of the event window of all airlines in the same sample group. Figures are reported in percentage rate terms. * Statistical significances at the 10% level. ** Statistical significances at the 5% level. *** Statistical significances at the 1% level. **** Statistical significances at the 0.1% level.
Fig. 1. Post-crash CARs crash (solid line) vs. non-crash airlines (dotted line).
crash date and this loss is statistically significant. In contrast, Panel B shows that the non-crash airlines do not experience any value loss in their shares on the crash date. Fig. 1 demonstrates their CARs over the 25-day post-event window. Apparently, despite the wealth of the shareholders of both groups suffer following the aviation disasters, the shareholders of the crash airlines experience a larger short-term wealth loss than those of the non-crash airlines. Kaplanski and Levy (2010) claim that aviation disasters could cause negative sentiment due to bad moods and anxieties and may in turn affect the stock prices of airlines. We argue that the fatality reinforces such a negative atmosphere and hence the scale of fatality plays a critical role in determining the post-crash stock price movement of the airlines. Table 2 shows that the investors of the crash airlines suffer larger losses on the crash day as the number of people killed in the accidents increases. Specifically, the average abnormal return of the crash airlines on the event date for the single-digit disasters (Panel A) is as low as −0.85%
Fig. 2. Degree of fatality and cumulative abnormal returns to crash airlines.
compared to −1.40% for the double-digit disasters (Panel B) and −4.89% for the triple-digit disasters (Panel C). Fig. 2 reveals that the crashes with larger fatality cause greater losses of the equity value of the crash airlines. The negative post-crash cumulative return drops dramatically as time elapses for the disasters with tripledigit fatality but only mildly for the double-digit fatality events. Interestingly, the (cumulative) negative reactions to single-digit fatality crashes disappear one week after the events,3 suggesting that the market soon recovers its confidence in the crash airlines after a week if the fatality is minor.
3 The authors thank for the anonymous referee’s suggestion to explore the patterns in CARs in a longer post-event window.
116
J.C. Ho et al. / Economics Letters 118 (2013) 113–117
Table 3 Abnormal returns of non-crash airlines by fatality. Date
AR
Panel A: low fatalities (n = 465) 0 0.13 +1 0.11 +2 0.04 +3 0.05 +4 0.17 +5 0.00 +10 0.40 +15 0.05 +20 0.02 +25 0.09 Panel B: medium fatalities (n = 600) 0 −0.01 +1 −0.21 +2 −0.30 +3 −0.06 +4 0.05 +5 −0.20 +10 −0.22 +15 0.30 +20 −0.06 +25 −0.28 Panel C: high fatalities (n = 134) 0 −0.34 +1 −0.68 +2 0.28 +3 −0.11 +4 0.06 +5 0.00 +10 −0.28 +15 0.31 +20 −0.12 +25 0.37
Patell test
Sign test
Window
0.649 0.894 0.514 0.032 1.231 −0.169 2.466** −0.677 −0.190 1.179
1.232 0.210 −0.441 −0.441 −0.906 0.767 1.650* 1.512 0.675 0.303
(0, 0) (0, 1) (0, 2) (0, 3) (0, 4) (0, 5) (0, 10) (0, 15) (0, 20) (0, 25)
0.13 0.24 0.27 0.33 0.50 0.50 1.27 1.56 1.53 1.25
−0.515 −2.075** −3.223*** −0.205
−0.337
(0, 0) (0, 1) (0, 2) (0, 3) (0, 4) (0, 5) (0, 10) (0, 15) (0, 20) (0, 25)
−0.01 −0.22 −0.52 −0.58 −0.54 −0.74 −1.08 −0.98 −1.45 −1.38
−0.515 −1.824* −3.360**** −3.028*** −2.585*** −3.388**** −3.531**** −2.725**** −3.377**** −2.796***
−0.337
(0, 0) (0, 1) (0, 2) (0, 3) (0, 4) (0, 5) (0, 10) (0, 15) (0, 20) (0, 25)
−0.34 −1.01 −0.73 −0.84 −0.77 −0.78 −0.57 −1.36 −1.62 −1.35
−1.536 −2.701*** −1.860* −1.795* −1.640 −1.448 −0.909 −1.616 −1.614 −0.930
−1.664* −2.876*** −1.491 −1.838* −2.011** −0.799 −0.106 −0.626 −1.318 −1.491
−0.741 −2.376**
0.154 −3.119*** 0.236 0.563 −2.055** −1.974** 1.626 −1.238 −3.447****
−1.536 −2.271**
−1.664* −3.050***
0.160
−2.568** −1.859* 3.220***
0.593
0.413
−0.412 −0.066
−1.664* −1.318
0.102
−0.888 1.149
0.413
−2.184** 2.318**
−0.952
−0.279
1.945*
0.933
CAR
Patell test
Sign test
0.649 1.089 1.196 1.094 1.548 1.322 2.416** 2.037** 1.864* 1.246
1.232 0.024 0.582 0.210 −0.069 1.418 2.906*** 1.605 2.348** 1.976**
0.236
−2.055** −1.810* −1.401 −3.446**** −2.874*** −3.365**** −2.710*** −1.156
Crashes leading to fewer than 10 fatalities are included in Low Fatalities. Crashes that have caused between 10 and 99 deaths are classified as Medium Fatalities. The High Fatalities group contains crashes that have killed more than 100 people. AR is the average daily abnormal return of all airline companies in the same group. CAR is the cumulative average abnormal return of the event window of all airlines in the same sample group. Figures are reported in percentage rate terms. * Statistical significances at the 10% level. ** Statistical significances at the 5% level. *** Statistical significances at the 1% level. **** Statistical significances at the 0.1% level.
Fig. 3. Degree of fatality and cumulative abnormal returns to non-crash airlines.
Fig. 3 shows that the direction of the impact of the aviation disasters on the stock prices of the non-crash airlines is determined by the relative importance of the ‘‘switch’’ effect and the ‘‘contagion’’ effect. We expect that the ‘‘switch’’ effect dominates the ‘‘contagion’’ effect when the fatality is minor, while the ‘‘contagion’’ effect begins to dominate the ‘‘switch’’ effect as the fatality increases. In the former scenario, the non-crash airlines may benefit from the misfortune of the crash airlines because passengers may switch to those non-crash airlines following aviation disasters. In the later scenario where large-scale fatality occurs, people may just avoid air travel after the disasters. As a result, the tragic news creates a contagion of fear that spreads through the entire airline industry.
Table 3 presents our test of this hypothesis. Panel A shows that both the ARs and CARs for the low-fatality air crashes during the post-crash period are consistently positive, suggesting that investors perceive that a small-fatality accident would benefit the non-crash airlines because passengers of the crash airlines may switch to the non-crash airlines and hence increase the cash flows of the rivals of the crash airlines. On the contrary, as the number of people killed exceeds ten, investors predict that such relatively large tragedies would cause a comprehensive bad sentiment concerning air travel safety among the public and potential increased operating costs for the entire airline industry. Consequently, even the non-crash airlines are adversely affected by the misfortune of the crash airlines. The consistently negative CARs of the non-crash airline reported in Panels B and C clearly provide supportive evidence for our argument that the ‘‘contagion’’ effect prevails in large-scale disasters. Overall, consistent with the findings of Kaplanski and Levy (2010) that large-scale aviation disasters affect people’s moods and increase their anxieties and consequently reduce their investment in risky assets, our results indicate that the more disastrous the aviation accidents, the stronger the fears and anxieties about air travel that impact the stock market. Finally, we examine whether there exists a difference between the means of the post-crash abnormal cumulative returns between each pair of fatality groups of stocks. Table 4 shows that the postcrash stock performance of both the crash and non-crash airlines is more significantly negatively affected by the high-fatality disasters
J.C. Ho et al. / Economics Letters 118 (2013) 113–117 Table 4 Comparisons of post-crash abnormal returns between different fatality groups. Event window
Low–medium
Medium–high
Panel A: comparisons between crash airline groups (0 , 0 ) 0.85 1.32 (0 , 1 ) 2.13** 2.44** (0 , 2 ) 3.07*** 0.49 (0, 3) 3.09*** 0.55 (0, 4) 3.05*** 0.46 (0, 5) 3.39*** 0.07 (0, 10) 3.17* 6.06* (0, 15) 2.90 8.13* (0, 20) 4.09 8.05 (0, 25) 2.79 11.34 Panel B: comparisons between non-crash airline groups (0, 0) 1.06 1.62 (0, 1) 1.68* 2.32** (0, 2) 1.17 2.06** (0, 3) 0.83 1.91* (0, 4) 1.49 2.09** (0, 5) 1.19 2.14** (0, 10) 2.35*** −0.52 (0, 15) 2.45*** 0.32 (0, 20) 2.98*** 0.18 *** (0, 25) 2.47 −0.46
Low–high 1.81* 3.70*** 2.23** 2.43** 2.34** 2.22** 9.23** 11.03** 12.13** 14.13* 1.85* 2.83*** 2.44** 2.20** 2.67*** 2.49** 1.84** 2.77*** 3.16*** 2.01
This table reports whether two fatality groups have an equal mean of cumulative abnormal return in an event-window. For both the crash airline and the non-crash airline sample, Low, Medium, and High contain all observations relating to air crashes that have killed fewer than 10 people, between 10 and 99 people and over 100 people, respectively. The t-statistics of the tests are reported followed by the indications of statistical significance levels. * Statistical significances at the 10% level. ** Statistical significances at the 5% level. *** Statistical significances at the 1% level.
than by the low-fatality accidents. Similar findings, though less evident, also exist between the low- and medium- fatality groups as well as the medium- and high-fatality groups. These results indicate that the market responds asymmetrically to the aviation disasters with various degrees of fatality. 4. Concluding remarks We examine the stock price performance of the airline industry following aviation disasters that have caused different levels of fatality. The results show that stock investors perceive differently the degree of the severity that aviation disasters may influence the stock performance of the crash and non-crash airlines. The crash airlines suffer larger losses of the equity value than
117
their competitors when aviation disasters occur. When aviation disasters result in single-digit fatality, the stock prices of the crash airlines initially respond negatively to the tragedy but recover quickly a week later. A higher degree of fatality has a larger and more persistent negative impact on the stock prices of the crash airlines following air crashes. Our explanation is that aviation disasters cause considerable financial damage to the crash airlines due to the reduction in air travel demand, possible future higher insurance costs, reputation and brand name damage, and customer goodwill loss. For the non-crash airlines, the ‘‘switch’’ effect dominates if the number of deaths in an accident is lower than 10 as these airlines benefit from the misfortune of the crash airlines. At higher fatality levels, however, the ‘‘contagion’’ effect dominates, as suggested by the declines of the stock prices following more serious crashes. Our explanation for the dominance of the ‘‘switch’’ effect for smallscale aviation disasters is that low fatality crashes receive less attention from the public and concerns about air travel safety. Air travelers may simply switch to the non-crash airlines to meet their travel needs after the occurrence of the disasters. On the contrary, large-scale air crashes tend to attract more attention from the media. Intensive and constant follow-up reports in the media for such devastating disasters reinforce greater fears and anxieties from the public than less serious ones. As a result, the non-crash airlines also suffer as the bad news creates a contagion of fear of air travelling that spreads through the public. References Borenstein, S., Zimmerman, M.B., 1988. Market incentives for safe commercial airline operation. The American Economic Review 78 (5), 913–935. Bosch, J.C., Eckard, E.W., Singal, V., 1998. The competitive impact of air crashes: stock market evidence. Journal of Law and Economics 41 (2), 503–519. Chance, D.M., Ferris, S.P., 1987. The effect of aviation disasters on the air transport industry. Journal of Transport Economics and Policy 31 (2), 151–165. Gigerenzer, G., 2004. Dread risk, September 11, and fatal traffic accidents. Psychological Science 15 (4), 286–287. Kaplanski, G., Levy, H., 2010. Sentiment and stock prices: the case of aviation disasters. Journal of Financial Economics 95 (2), 174–201. Mitchell, J., 2001. Clustering and psychological barriers: the importance of numbers. The Journal of Futures Markets 21 (5), 395–428. Mitchell, M.L., Maloney, M.T., 1989. Crisis in the cockpit? The role of market forces in promoting air safety. Journal of Law and Economics 32 (2), 329–355. Silver, R.C., Holman, E.A., McIntosh, D.N., Poulin, M., Gil-Rivas, V., 2002. Nationwide longitudinal study of psychological responses to September 11. The Journal of the American Medical Association 288 (10), 1235–1244. Sivak, M., Flannagan, M.J., 2004. Consequences for road traffic fatalities of the reduction in flying following September 11, 2001. Transportation Research 7 (4–5), 301–305.